Americas Churning Races:

Race Response Changes between Census 2000 and Census 2010

Carolyn A. Liebler, University of Minnesota
James Noon, U. S. Census Bureau



Paper proposal for the 2014 meetings of the Population Association of America

DRAFT: DO NOT CITE OR QUOTE WITHOUT PERMISSION
Disclaimer: This report is released to inform interested parties of research and to encourage discussion. The views
expressed are those of the authors and not necessarily those of the U.S. Census Bureau.



Brief (150 word) Abstract:
Race responses, Hispanic responses, and corresponding identities can change over time
and across contexts. Are these changes widespread or rare? Which groups are affected by such
changes? We use internal Census Bureau data from the 2000 and 2010 censuses in which
individuals responses have been linked across years to answer these questions. We find that
millions of individuals (about 6% of people in our sample) changed their race and/or Hispanic
origin responses between 2000 and 2010. These changes occur in every direction into and out
of single race groups, Hispanic and non-Hispanic designations, and across multiple-race
categories. Response changes are widespread, happening among males and females, youth and
adults, in the West region and in other regions. Multiple-race response groups experienced
substantial churning, as did the race responses of Hispanics, American Indians, and Pacific
Islanders. Some groups show substantial stability in race/Hispanic responses, particularly single-
race non-Hispanic whites, blacks, and Asians.

1

Race responses, Hispanic responses, and corresponding identities sometimes change over
time and across contexts (Alba and Islam 2009; Duncan and Trejo 2011; Eschbach 1993; Harris
and Sim 2002; Perez and Hirschman 2009; Saperstein and Penner 2012). Are these changes
widespread or rare? Which groups are affected by such changes? Using remarkably well-suited
data, we document the extent to which individuals change their race and/or Hispanic origin
responses between Census 2000 and Census 2010.

Prior Knowledge and Hypotheses
Future versions of this paper will have a substantial review of related literature aimed at
formulating hypotheses about what we might find in these analyses.

Data
The U.S. Census Bureau has endeavored to link individuals census records as part of an
effort to understand response variability and reduce future data collection costs. We use internal
Census Bureau data from the 2000 and 2010 censuses in which individuals responses have been
linked across years by the Census Bureaus Center for Administrative Records Research and
Applications (CARRA). CARRA uses probability record linkage techniques and personal
information such as name and date of birth to assign a Personal Identification Key or PIK (see
Wagner and Layne 2013). This PIK is used to link a persons record in one census or survey (in
this case, Census 2000) to their own record in another census or survey (in this case, Census
2010). CARRA anonymizes the linked data so that it can be used for Census Bureau statistical
purposes and for research such as this study. Improvements to linkage data and methods are
ongoing, and we use the best available data. Naturally, a few PIKs are not assigned to the correct
2

person. An assessment of false match rates yielded estimates ranging from 0.2% to 6.6% (Layne
et al. 2013); we conservatively assume a false match rate of 6.6% in the results presented below.
The linked data in our study naturally exclude people who answered Census 2000 but
died or left the country by 2010. Similarly, new immigrants and children under age 10 in 2010
are not included in the linked data. The linked data also exclude people who do not have a social
security number or individual tax identification number (e.g., foreign guests) and those whose
personal information was too ambiguous or incomplete to assign a PIK. Our primary data source
contains about 193 million individuals who were present and able to be assigned a PIK in both
the 2000 and 2010 full-count decennial censuses.
For several reasons, these data are remarkably well-suited for a study of race and
Hispanic response change over time. First, the race and Hispanic origin questions did not
undergo major changes over the decade and so many changes in the responses can be attributed
to other factors. Second, the data have a very large sample size and cover a considerable portion
of the US population. Third, because of the large sample size, we can examine movement
between over 100 race/Hispanic origin categories in each census. And fourth, these are the
decennial census data data commonly used to study Americas demographics and undergirding
extensive social science research and policy analysis on racial disparities and how the disparities
are changing over time.

Because we are using full-count data data collected on the short form versions of each
census we are limited in what we know about individuals. We have three hypotheses based on
prior research:
H1: Women are socialized to have more complex and nuanced identities than men, and
so women will be more likely to change their race/Hispanic responses.
3


H2: Young people (ages 0-17 in 2000) will be more likely to change their race/Hispanic
responses for three reasons. First, they were probably reported by their parents in 2000
but may be self-reporting in 2010. Second, childhood and adolescence are times of
personal identity development while adulthood involves a more stable identity. And third,
many children of interracial/interethnic unions will still be children in 2000 while a lower
proportion of adults in 2000 is thought to have strong ties to multiple race/Hispanic
groups.

H3: People in the west, especially California and Hawaii, have a tradition of accepting
complex and changing racial/Hispanic identities, so there will be more change in the west
region than in other regions.


Conclusion
We will use matched census records for about 197 million Americans who in the 2000 and 2010
decennial censuses, and focus on 166 million who were most likely to self-report (or have a
household member report) their race/Hispanic origins. Using these remarkable data, we will give
a first look into the extent of race/Hispanic response churning happening in contemporary
America. We see our work as having substantial implications for sociological theories about race
and Hispanic identities, as well as for research on racial stratification and inequalities. In future
versions of this manuscript, we will expand on both of these points.


4

References
Alba, R., & Islam, T. (2009). The case of the disappearing Mexican Americans: An ethnic-
identity mystery. Population Research and Policy Review, 28, 109-121.
Duncan, B., & Trejo, S. (2011). Who remains Mexican? Selective ethnic attrition and the
intergenerational progress of Mexican Americans. In D. Leal & S. Trejo (Eds.), Latinos and
the economy (pp. 285-320). New York: Springer.
Eschbach, Karl. (1993). Changing identification among American Indians and Alaska Natives.
Demography, 30 (4), 635-652.
Harris, D., & Sim, J.J. (2002). Who is multiracial? Assessing the complexity of lived race.
American Sociological Review, 67 (4), 614-627.
Humes, K., Jones, N., & Ramirez, R. (2011). Overview of race and Hispanic origin: 2010. 2010
Census Brief C2010BR-02. Washington, DC: U.S. Census Bureau.
Layne, M., Wagner, D., and Rothhaas, C. (2013). Estimating Record Linkage False Match Rate for the
Person Identification Validation System. CARRA Research Report Series (CARRA0613). Center
for Administrative Records Research and Applications, U.S. Census Bureau.
Perez, A., & Hirschman, C. (2009). Estimating net interracial mobility in the United States: A
residual methods approach. Sociological Methodology, 39, 31-71.
Rastogi, S., Johnson, T., Hoeffel, E., & Drewery, M. (2011). The Black Population: 2010. 2010
Census Brief C2010BR-06. Washington DC: U.S. Census Bureau.
Saperstein, A. and Penner, A.M. (2012). Racial fluidity and inequality in the United States.
American Journal of Sociology, 118(3), 676-727.
Wagner, D., and Layne, M. 2013. The Person Identification Validation System (PVS): Applying the
Center for Administrative Records Research and Applications (CARRA) Record Linkage Software.
CARRA Research Report Series (CARRA0513). Center for Administrative Records Research and
Applications, U.S. Census Bureau.